Feature Selection of Technical Indicators Using Mutual Information and HSIC for Financial Time Series Analysis.

In this paper, we present a feature selection technique for the NIFTY50 index using the technical indicators. Two nonlinearly related measures, namely Mutual Information (MI) and the Hilbert--Schmidt Independence Criterion (HSIC), are employed to rank the predictive capabilities of 20 technical indicators. The methodology utilizes adjusted historical market data, which is processed using a combination of technical analysis indicators, including moving averages, oscillators, and volume-based indicators. The results indicate that the feature rankings obtained using MI and HSIC exhibit noticeable differences. According to MI, the most informative indicators are the Ichimoku Conversion Line (3.250), WMA\_14 (3.222), and EMA\_14 (3.174). In contrast, HSIC identifies SuperTrend (3197.847), Ichimoku Conversion Line (2827.923), and WMA\_14 (2827.892) as the most significant features. The proposed feature selection framework supports data-driven financial decision-making and contributes to financial market efficiency, aligning with the United Nations Sustainable Development Goal 8.